Concurrency Limits

TriggerFlow has three commonly used concurrency control layers:

• batch(..., concurrency=n)
• for_each(concurrency=n)
• create_execution(concurrency=n) / execution.set_concurrency(n)

Recommended mental model

• batch / for_each limit local fan-out
• execution-level concurrency protects the whole execution globally

Recommended example

execution = flow.create_execution(concurrency=2)
execution.set_concurrency(1)

Current best practices

• when you have API or DB rate limits, prefer an execution-level global cap
• when you only need local throttling, use batch / for_each concurrency
• in most service deployments, using both layers is the most robust approach

No longer recommended

• letting one execution saturate all external dependencies without limits
• using concurrency settings to compensate for a bad orchestration topology